# Finite Element Analysis Results In Smart Infill

If you would like to make a 3D print stronger, just add more material. Increase the density of the infill, or add more perimeters. The problem you’ll encounter though is that you don’t need to add more plastic everywhere, only in the weak areas of the part that will be subjected to the most stress. Studying where parts will be the weakest is the domain of finite element analysis, and yes, you can do it in Fusion 360. With the right techniques, you can make a stronger part on your 3D printer, and [Stefan] is here to show you how to do it.

The inspiration for this build comes from [Adrian Bowyer]’s blog, where he talks about adding ‘fibers’ to the interior of 3D printed objects to increase strength. These ‘fibers’ aren’t really fibers at all, but long, thin, cylindrical voids. The theory of this is that the slicer will interpret this as a hole and place perimeters around these voids, effectively increasing the density of the infill in a local area in the print. Combine this with finite element analysis, and you get a part that is stronger where it needs to be, and doesn’t waste plastic.

However, there is an easier way. Fusion 360 and ANSYS Finite Element Simulation are both free-ish tools that allow for some amount of finite element analysis on an imported 3D object. This can be used to find the weakest part of any 3D print, and it can this can be exported as a 3D mesh. Slic3r has a modifier mesh function, and combining this finite element analysis mesh (printed at 100% infill) with the original part (printed at 10% or so infill) results in something that’s strong where it needs to be, doesn’t waste plastic, and is much easier to set up than [Adrian Bowyer]’s ‘fiber’ technique.

After printing a few 3D printed hooks with varying degrees and techniques of infill, [Stefan] found the baseline of 2 perimeters failed in a test hook at about 50kg load. The Smart Infill hook failed at about 100kg. Not bad, and the fancy-pants hook only weighs about 30% more.

You can check out a video of the entire toolchain and testing below. Thanks [Keith] for sending this one in.

## 16 thoughts on “Finite Element Analysis Results In Smart Infill”

1. You can do something similar if designing your parts in OpenSCAD (or I imagine any CAD) with a for loop If you need strength in an area, drop in 1,2, or 3 extrusion width ribs exactly 4,5,or 6 extrusion widths apart. You can even more or less choose the contiguous paths the slicer picks in these areas by controlling the extrusion width count. I use these to support huge bridges as well at the 1 extrusion width level in a format a lot easier to cleanup than traditional supports.

1. elwing says:

Doesn’t solve the finite element part, but yes, thanks for the tip

2. Nathan McCorkle says:

very cool, now we just need to hook this up with Fenics/DOLFIN for a totally open-source approach! (other suggestions are appreciated too)

1. Foyle says:

3. Nate B says:

I think this is finally the realization of one of 3d printing’s greatest potentials. It’s just yet another article on Hackaday, but I think in ten or twenty years, people will be looking back at this moment as a turning point in an industry, moreso than most other articles here.

1. I’m actually with you on this one. I was just re-reading and old article (https://hackaday.com/2016/04/07/a-look-into-the-future-of-slicing/) and thinking “what _is_ the state of FEA modelling in 3DP?” This is a partial answer. :)

I’ll admit that when I need strength, I usually go with the “5 perimeters” method myself, but this is much more nuanced.

I need to play around with Slic3r’s modifiers a lot more.

And BTW: the basic fibers trick is an old one. I have no idea where I learned it from, but “drilling” holes through a part to trick the slicer into making perimeters around it is a great one to have in your basket. I’ve used it a bunch — manually of course. Combining the fibers trick with the output of an FEA program actually sounds great to me: I don’t know why the author of the video dismisses it so quickly.

2. Mike says:

“but I think in ten or twenty years, people will be looking back at this moment as a turning point in an industry”

I respectfully disagree. Unfortunately the progress in the industry is slowing to a crawl and make no mistake about it, it is intentional. For example, take a look at the article linked in the comment above mine, especially in the bottom edit portion where STEP has been defined for much more than being used today. This is because software manufacturers prefer to use proprietary file formats and now some don’t even let you save your own files (unless its in a geometry only format with no design intent stored) like fusion 360. This lack of interoperability is what slows growth as those companies have no motivation to really increase their features or tool chains, instead they focus on customer lock in (like cloud saving without file portability).

Features like this should have been implemented in CAD products years ago, just like slicers should have been implemented shortly after 3d printers started selling in volume. If file interoperability was actually a thing then we would have multiple competing software packages that would include the entire tool chain from idea to output, rather than have to find ways to try and get several software packages to work together like the 360 to ansys to slic3r as shown here.

4. DougM says:

Unfortunately the part in the Z axis can only be as strong as the between-layer bonding strength of your material. Usually you can get around this by orienting your part, but sometimes you can’t. Especially in my case where my printer is not good at all at printing supports.

There is, so far as I can tell, no reason why 3D printers have to lay down material layer by layer. Moving the Z axis up and down in conjunction with the movement of the XY axes is just math.

A properly trained AI (buzzword alert) could easily run finite element analysis on a million simulated strategies before starting your print.

1. Carlitos says:

“There is, so far as I can tell, no reason why 3D printers have to lay down material layer by layer.”
…because the extruder will hit already placed material otherwise?

2. This (https://hackaday.com/2016/07/27/3d-printering-non-planar-layer-fdm/) is an epic Hackaday article on non-planar 3DP. Has anyone pushed this idea further than Moritz did?

Noodle-based extrusion methods do have a weaker Z axis, but still, the strength in the Z direction goes up with surface area — so solid infill should provide a lot more strength there too.

Which is to say, this method still applies, but you’d certainly want your FEA program to know something about the material’s “grain”.

Is this an AI problem? AI is a statistical method that’s great when you’ve got a crap-ton of data, but you don’t really know how to model it. Here, we understand the physics pretty well, but there’s very little data on the printed results. I feel like this one just calls for some good old-fashioned hard thinking.

1. DougM says:

I’m sure you’re right about the hard thinking vs. AI. If you could simply enter as part of the build setup process the direction and strength of your stresses a good math engine could derive the best printing method or tell you if your part won’t survive no matter what it does.

However, if computing power is free then properly trained AI could potentially come up with solutions outside the bounds of your math engine.

5. I would like to see the results of the tests after they have been annealed. That usually increases the resistance to de-laminating and generally makes 3d printed parts more resilient.

6. I would be curious to know how does that compare to gyroid infill.

7. This is being done at a much larger scale, in a much more integrated and powerful way by many in the industry such as Autodesk. It’s called generative design, you can use an early version of it in Fusion360. We just give the bounding geometry, the loading conditions and values, and the software iterates on the geometry via a genetic algorithm and then uses integrated FEA to check the new design until it finds the optimum geometry.

This site uses Akismet to reduce spam. Learn how your comment data is processed.